TEST DUMMIES
IT’S US; we’re the test dummies. Except we’re not dummies, because all three of these builds booted on first attempt. Big pat on the back for us, round of applause, bullhorns, etc. We’re pretty happy with how the build process went, barring the mechanical difficulty of installing the necessary brackets on the Fractal Design cooler in the 4K system. No significant hiccups to discuss here, though.
Before we get into the nitty-gritty details of our test results, let’s break down the decision-making process behind our next selection of benchmarks. You’ll be seeing these tests a lot in the coming issues of MaximumPC, so we felt it best to address why we chose the benchmarks we chose.
First up, we’re keeping two bits of software from our old benchmark tables: CrystalDisk and 3DMark FireStrike. CrystalDisk needs no real introduction; it’s a super-reliable drive speed test, which gives us real-world figures for easy comparison. We’re still using version 5.2 of the program, which we’ve found to be reliable and avoids the feature bloat of later iterations. With FireStrike (and its 1440p and 4K variants, Extreme and Ultra), we opted to keep it since it’s an extremely common choice of graphics and physics test, and the base test is available for free in the trial version of 3DMark.
On to the new stuff. Cinebench remains the best option for straightforward CPU testing, so we’ve upgraded to the most recent version (Cinebench R23) and will be testing both single-core and multi-core performance going forward, for greater clarity. R23 does take longer to bench than R20 since it fills a ten-minute slot with as many passes as it can, but that’s a fair trade-off for greater test accuracy.
With games, we’ve expanded our selection to include some new game benchmarks. TotalWar:ThreeKingdoms replaces TotalWar:WarhammerII as our CPU-bound strategy title, perfect for measuring combined CPU and GPU performance. Elsewhere, we’ve got MetroExodus and Cyberpunk2077, the latter of which has a particularly brutal RTX benchmark that incorporates raytraced reflections, shadows, and general lighting effects. We’ve tested both RTX and regular graphical settings here for both of these games, to demonstrate the practical impact using ray-tracing will have on your gaming experience.
All of these games were tested at the ‘Ultra’ graphical preset (although it should be noted that MetroExodus features a GPU-hammering ‘Extreme’ option too) at native resolutions, with the appropriate version of 3DMark FireStrike used for each build. Crucially, we’re not using Nvidia’s DLSS or AMD’s FSR options; these are base performance results, and using these resolution-scaling options is almost guaranteed to result in better performance but also greater variation between GPU models. We’re using AMP/ XMP memory profiles for full RAM speed, but no manual CPU overclocking.
Benchmarking Results
Although we took the time to crossreference our results with other users’ findings online, we weren’t especially surprised by any of the figures we got here. Looking at the 1080p build first, we were pleased to see that the RX 6600 XT performed well at FHD resolution, nailing 60+fps in all of our non-ray-traced game tests and scoring highly in 3DMark FireStrike.
Turning on RTX settings did reveal that the 6600 XT isn’t well-suited to raytraced tasks, though. A significant hit took our framerate below 60fps in Metro Exodus, while the legendarily demanding Cyberpunk Ultra RTX benchmark came back with an unplayable eighteen frames per second. The Ryzen 5 5600X did well in TotalWar:ThreeKingdoms and returned a solid single-core score in Cinebench R23 but, unsurprisingly, lagged far behind the other two builds in multi-core performance.
SSD performance placed our 1080p build pretty far beneath the 1440p and 4K machines too, with the ADATA Legend 840 not hitting its advertized speeds (but still managing a respectable 3.3GB/s read and 2.4GB/s write in CrystalDisk). Meanwhile, the other machines sit close to the peak of M.2 Gen4 capabilities, with the Samsung 980 Pro lagging a little in write speeds while the MSI Spatium M480 drive actually exceeded its advertized speed of 7GB/s by a small amount, which was impressive.
In our 1440p rig, we saw solid performance across the board at our target resolution, with notably excellent single-core performance in Cinebench R23 from the Intel i7-12700K; most likely thanks to that CPU’s new performance/ efficiency core architecture. Accordingly, this build also excelled in TotalWar. Raytracing in games still caused a major performance dip, although, in practical terms, this could be mitigated by the use of DLSS.
Finally, the ultra machine did well, although struggled noticeably more with games at 4K. Cyberpunk2077 is renowned for battering graphics cards at higher resolutions, which was fairly demonstrated by these results, although the RTX 3090 handled 4K ray-tracing a little better than the RX 6600 XT did at 1080p. The Ryzen 9 5950X also didn’t quite measure up as well as we’d hoped, with single-core performance that fell behind the cheaper i7-12700K and lower framerates in the TotalWar benchmark compared to other games.
Closing Remarks
Our work here is done; we now have a shiny new set of benchmark results to compare all our future builds against. We’ll also be using the relevant set of scores for PC and laptop reviews going forward, although you’ll probably notice that those figures aren’t present in our reviews section this issue—time constraints and different review writers made this impossible.
Building and shooting three PCs for one feature was certainly a challenge, and not one we’ll be repeating anytime soon. But with our updated benchmarks, we’re hopeful that we’ll be able to provide a good reference point for our future builds for at least the next year or two. But before long, they’ll need updating again, and we’ll be ready to rise to the challenge.